[lld] [lld][AMDGPU] Support R_AMDGPU_ABS32_(LO|HI) relocations (PR #191597)

[Joseph Huber via llvm-commits]

================
@@ -0,0 +1,22 @@
+# REQUIRES: amdgpu
+
+# RUN: llvm-mc -filetype=obj -triple=amdgcn-- -mcpu=fiji %s -o %t.o
+# RUN: ld.lld -shared -Bsymbolic %t.o -o %t
----------------
jhuber6 wrote:

No, we don't pass it by default. This was just me being lazy when I should've just marked it as protected or hidden to prevent those errors. In practice AMDGCN emits hidden for most things and protected for things exported to the runtime.

https://github.com/llvm/llvm-project/pull/191597